The invention relates to a power plant coal storage amount short-term prediction method based on SSA and LSTM deep learning. The method comprises the following steps of acquiring historical data and temperature data of the power plant coal storage amount, the coal feeding amount, the coal consumption amount and the power generation amount of a set number of days before a prediction time node to serve as input data; carrying out embedding, decomposition, recombination, reconstruction and normalization processing on the input data by adopting an SSA algorithm; inputting the processed data into atrained LSTM network model to predict the daily coal storage amount of the power plant for a set number of days after the time node is predicted; and carrying out inverse normalization on the prediction result and outputting a result. Compared with the prior art, the accuracy of short-term prediction of the daily coal storage amount of the power plant is effectively improved.